Colour_Reproduction_in_Electronic_Imaging_Systems_Photography_Television_Cinematography_2016_Michael_S_Tooms

Colour in Cinematic Production – The Academy Color Encoding System 573 CIE 1931 x,y chromaticity diagram 1.1 G 540 550 1.0 ACES 0.9 520 0.8 530 510 0.7 y 0.6 560 570 500 580 0.5 0.4 D60 590 0.3 490 EE White 0.2 600 610 620 640 700 R 480 0.1 470 0.0 460 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 4000.2 –0.1 B –0.2 x Figure 32.3a The chromaticity gamut of the ACES colour space on the CIE x,y chromaticity diagram. It will be noted that the blue primary is not located precisely at the point indicated by the criteria specified at the beginning of this section; that is, it is located slightly off the x-axis with a value of x = 0.0001 rather than zero. The reason for this ‘offset’ appears to be somewhat arcane, with the author failing to discover the basis for it, despite email correspondence with two of the IIF Subcommittee members. The system white point appearing in Table 32.1 is daylight, specified as the CIE Illuminant D60 and appears to have been selected as the nearest CIE illuminant to the average chromaticity of the xenon-based projectors used for theatre reproduction. This places the ACES system white between the photographic system white of D50 and the television system white of D65. 32.3.2 Relative Luminance Contrast Range In Section 13.3, a dynamic range of 5,000:1 was suggested as approaching the limit of the contrast range of the eye; thus, if the ACES colour space is to accommodate the full dynamic

574 Colour Reproduction in Electronic Imaging Systems CIE 1976 u′, v′ chromaticity diagram 0.7 0.6 G 570 520 530 540550 560 580 590 510 600 0.5 500 D60 EE white 610 620 630 640 660 700 R 0.4 490 v′ 0.3 480 0.2 470 0.1 460 450 440 400 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 –0.1 ACES –0.2 –0.3 B –0.4 u′ Figure 32.3b The chromaticity gamut of the ACES colour space on the CIE u′,v′ chromaticity diagram. range the eye is capable of perceiving, then it would be wise to regard this range as the minimum to be embraced. The ACES colour space is scene referred; that is, it is a linear colour space, which means it does not have the protection from digital contouring artefacts afforded by perceptual uni- form coding. By using the relationships defined in Section 13.6 as the criteria to establish the number of bits required to ensure that no digital contouring of the image is perceived, we can use Worksheet 13(c) to obtain the graphs illustrated in Figure 32.4, which show the perceptibility of bit contouring against the human visual modulation threshold (HVMT) over the luminance range of the display using a binary fixed-point 16-bit digital encoding regime.

Colour in Cinematic Production – The Academy Color Encoding System 575 Perception of quantisation contouring - 16 bit simple encoding 1 0.1 Weber–Barten x 10 Modulation 0.01 Weber–Barten 1% Weber–Barten HVT 0.001 γ =1 0.0001 Display contrast range 1000 0.001 0.01 0.1 1 10 100 (a) Display luminance in nits or (cd/m^2) Perception of quantisation contouring - 16 bit simple encoding 1 0.1 Weber–Barten x 10 Modulation 0.01 Weber–Barten 1% Weber–Barten HVT 0.001 γ =1 0.0001 Display contrast range 1000 0.001 0.01 0.1 1 10 100 (b) Display luminance in nits or (cd/m^2) Figure 32.4 Graphs illustrating the equivalent modulation level of a single binary bit change with a display contrast range of 5,000:1. In the upper graph, the display highlight is set to 100 nits and in the lower graph to 1,000 nits.

576 Colour Reproduction in Electronic Imaging Systems From the upper graph, it can be seen that the line representing the modulation caused by a linear (������ = 1) encoding crosses the HVT at a display luminance level of 0.2 nits, that is, at a contrast range of 500:1. If the screen highlight level is increased to 1,000 nits, as illustrated in the lower graph, then the crossover occurs at about 4.0 nits, representing a perceived contrast ratio of only 250:1. The reduction is due to the luminance range being moved to the point on the HVMT curve where the eye is more sensitive to changes in modulation. Thus, on the basis of the criterion of requiring at least a contrast ratio of 5,000:1, 16-bit fixed-point coding is inadequate for the task. In order to accommodate the higher contrast ranges required, the ACES specifies a 16-bit half-precision floating-point format for coding the ACES RGB signals. This coding format is specified in the IEEE 754-20083 standard, where it is officially referred to as ‘binary16’ and is based upon a scheme originally defined by Nvidia to accommodate large contrast range image files. By splitting the available 16 bits into three groups, one bit for sign, five bits for the exponent and 10 bits for the fraction or mantissa, a wide dynamic range is achieved at the cost of a level of accuracy, which is nevertheless higher than that required for colour processing. As the footnote reference indicates, four bits are effectively available for the exponent, which thus has a maximum value of 24 or 16: an exponent of 16 applied to a base of 2 will provide for up to 216, that is, 65,536 values. However, as the reference shows by example, the maximum value achieved by the coding method is actually 65,504, and since there is a sign bit, this equates to an availability of levels between −65,504 and +65,504. This method of coding will ensure that any level of signal will be encoded and decoded to a level of accuracy to a minimum of five decimal figures; thus, effectively the accuracy of the level of input signal is retained; there is no perceptible change in level between the original and the decoded values and thus no contouring of the rendered image. In addition, since this form of coding is effectively transparent in terms of input and output levels, there is no requirement to map the levels representing ‘black’ and ‘white’ to specific coding levels as there is when digitally coding television and photographic signals. The ACES is scaled such that a perfect reflecting diffuser under a particular illuminant produces ACES RGB values of 1.0. Many scenes include objects with radiance values greater than that of a perfect reflecting diffuser; hence, ACES values well above 1.0 can be expected. 32.4 Reference Input Capture Device (RICD) 32.4.1 RICD Spectral Sensitivities The reference input capture device illustrated in Figure 32.2 is a hypothetical device with ideal characteristics to capture the scene RGB values in terms of the ACES colour space characteristics. Thus, the RICD scene colour spectral sensitivities or spectral sensitivities in terms of the standard CIE XYZ colour matching functions (CMFs) can be found using the relationship derived in Appendix F and applied in Worksheet 32(a) which produces the matrix illustrated in Table 32.2. In Worksheet 32(a), the matrix illustrated in Table 32.2 is used to produce the RICD scene colour spectral sensitivities from the CIE CMFs; these spectral sensitivities are illustrated in Figure 32.5. 3 This format is described at http://en.wikipedia.org/wiki/Half-precision_floating-point_format

Colour in Cinematic Production – The Academy Color Encoding System 577 Table 32.2 The matrix for deriving the ACES RGB capture spectral sensitivities from the XYZ colour matching functions (CMFs)4,5 XYZ R = 1.0760118426 0.0000000000 −0.0000999175 G = −0.5082796015 1.4075877241 0.1006918774 B = 0.0000000000 0.0000000000 1.0159913478 1.6Relative response 1.4 1.2 1.0 0.8 ACES 0.6 0.4 0.2 0.0 380 420 460 500 540 580 620 660 700 740 –0.2 Wavelength (nm) Figure 32.5 The spectral sensitivities of the Reference Input Capture Device (RICD). Since the green and blue ACES primaries either match or are located close to the CIE Y and Z primaries and the red primary is on the extreme of the spectrum locus, the resulting spectral sensitivities have a marked similarity to the CIE CMFs illustrated in Figure 4.4. However, the ACES minor red lobe is slightly smaller than the equivalent x lobe due to the red primary being located well within the XYZ gamut. In Worksheet 32(a), the values appearing in WS Table 1 are convolved with the spectral power distribution (SPD) of CIE Illuminant D60 in WS Table 3 and the results for each convolution summed to show that equal values of RGB signal are obtained, that is, a white 4 The values in the matrix are given to 10 decimal places to be consistent with the accuracy given for the values in the ACES specification. However, there is a scaling difference between the figures in Table 32.2 and those in the specification because the worksheet normalises the G coefficients by adjusting the summation of the centre line coefficients of the matrix to be equal to 1.0. 5 If the criteria for locating the primaries on the CIE x,y diagram had been adhered to, the Z value of the red primary would have also been zero, making the matrix interestingly symmetrical, with four of the nine values equal to zero, the inevitable result of locating two of the primaries on the y-axis of the chromaticity chart.

578 Colour Reproduction in Electronic Imaging Systems balance for a scene neutral illuminated by a D60 source is achieved by the RICD with no balance adjustments being required. Table 32.3 Matrix for deriving XYZ values from RGB values RG B X = 0.9525523959 0.0000000000 0.0000936786 Y = 0.3439664498 0.7281660966 −0.0721325464 Z = 0.0000000000 0.0000000000 1.0088251844 In Worksheet 32(a) the matrix in Table 32.2 is inverted and the resulting Y coefficients are normalised to produce the matrix in Table 32.3, which expresses the XYZ values in terms of the RGB values. In contrast to the figures appearing in Table 32.2, these figures correspond precisely with those appearing in the ACES specification. 32.4.2 RICD Flare The RICD is defined as being free of capture system noise and to introduce flare amounting to 0.5% of the captured values of a perfect reflecting diffuser. The flare was specified in this manner to match the simplest plausible model of camera flare that could be used with the reference rendering transform (RRT) and produce a visually pleasing result. Moreover, it assumes that this level of camera flare is constant across the image plane and is constant for all values of captured content. 32.5 The Input Device Transform Since it is unlikely that the native spectral sensitivities of a real camera can be made to match those of the RICD, the purpose of the input device transform (IDT) is to transform the colour space of the real camera to that of the RICD. As described in the previous section, the ideal characteristics of the capture device are the spectral sensitivities defined by the CMFs of the system primaries, and it was shown in Section 17.2 that there are practical difficulties in shaping the characteristics of the optical colour filters and the image sensor(s) to match precisely the shape of any ideal set of CMFs and the same is equally true of the ACES primaries CMFs, as illustrated in Figure 32.5. If it were assumed that the capture device had responses approaching the shape of the blue and green CMFs and the primary response of the red CMF, then the secondary red response in the 380–500 nm range could be emulated by inverting a fraction of the blue signal and adding it to the red signal via a suitable matrix. A simpler approach would be to aim for the characteristics of an ideal camera response as described in Section 12.4, where the input device gamut is defined by selecting specific locations of primaries on the u′,v′ diagram, which leads, in the two examples given, to diminishing amplitudes of the red secondary lobe required. In the second of the examples, the secondary lobe is fast approaching zero level amplitude; thus, with such spectral sensitivities only a very low level of matrix coefficients would be required.

Colour in Cinematic Production – The Academy Color Encoding System 579 Once the optimum combination of spectral sensitivities and matrixing has been determined, any further improvement in matching the ideal curves would be achieved by establishing suitable look-up tables as discussed in Section 23.4.2. The Academy has produced a draft procedure, ‘P-2013-001 Recommended Procedures for the Creation and Use of Digital Camera System Input Device Transforms (IDTs)’, to provide guidance on the design of the IDT. 32.6 An IIF System Configuration for Viewing the Graded Signals Defined in the ACES Colour Space In order to avoid much repetition, it is assumed that the reader is familiar with the material related to appraising the rendered image contained in the first three sections of Chapter 26 and with the concepts of applying transforms to encoded images to match them to a reference viewing environment, as discussed in Sections 27.3.1.4 and 27.3.1.5. As noted earlier, it is intended that the ACES colour space is the adopted working colour space throughout the post-processes; however, it is evident that during processing and subse- quently, these processed signals will require transformation to a colour space or spaces matched to their display and its environment, in particular, the grading display and the cinema projector. It was also noted in Section 32.2 that the colour gamuts for both the post and the distribution workflows would encompass all colours; thus, when considering the required transformations, it is well to keep in mind that both these displays will be required to accommodate signals encoded in a colour space whose chromaticity coordinates extend beyond the spectrum locus. 32.6.1 Objective versus Preferred Rendition Assuming for the moment a display with a fundamentally linear transfer characteristic, then in colorimetric terms, a simple matrix to transform from ACES RGB coding to the display RGB coding would ensure rendered images with minimal colorimetric errors. Such a solution would provide excellent results in terms of colour fidelity; however, that is not necessarily what is required, since the intent is for the material to be appreciated by a cinema audience inculcated into viewing excellent-quality pictures resulting from the experience of decades of finessing the images derived from film with its logarithmic characteristics. The means of addressing the ACES method of extracting the essence of the desirable characteristics of the traditional film path characteristics and applying it to the ACES master is dealt with in the next section on the RRT. 32.6.2 Matching Reference and Cinema Displays A representative workflow of an ACES system is illustrated in Figure 32.6. The Digital Cinema Initiative (DCI) have adopted a naming regime for the master file prepared by post for distribution as the DSM and the format of the file used for distribution as the digital cinema distribution master (DCDM); this nomenclature has been used in the figure, which, in order to close the grading reference display matching loop, also includes the exhibition cinema and projector. A scene captured by the camera sensors and encoded in the native colour gamut of the camera is transformed by the IDT into the ACES colour space, which, with the exception of the displays, is the colour space then retained throughout the post-processes.

580 Colour Reproduction in Electronic Imaging Systems Display ODT Display ODT (RGB)1 ACES-(RGB)1 (RGB)2 DCDM-(RGB)2 Reference display Cinema display Colourist Reference Cinema room environment Grading controls environment Scene Camera Native IDT ACES Grading RRT ACES capture processor (OCES) Post Digital camera File File ODT File archive DSM ACES - DCDM DCDM master Figure 32.6 Matching graded material on the reference and cinema displays. It must be emphasised that the configuration illustrated in Figure 32.6 is simplistic; it is based upon a simple colorimetric approach where the display device is fundamentally of a linear nature6 and the problems of distributing the ACES 16-bit digital signal (McElavain et al., 2012) are not addressed. The grading loop includes the RRT (sometimes located in the reference display), which imposes desirable filmic characteristics on the image. It is defined in the specification as ‘the signal-processing transform that maps an ACES representation of an image to an Output Colour Encoding Space (OCES) representation appropriate for viewing on the Reference Display Device’. The OCES is defined as ‘the color image encoding used by the Image Interchange Framework to represent images to be displayed on the Reference Display Device (RDD)’; however, since the RDD is in turn defined as ‘an ideal output device with an unlimited color gamut and a dynamic range exceeding any current or anticipated real output device’, it would appear that the OCES is synonymous with the ACES colour space. The RRT encoding is transformed by the output device transform (ODT)7 in the display unit to the chromaticity gamut and the contrast range of the grading reference display. When considering the colour characteristics of the grading reference display, it might be assumed as an initial consideration that it should match the characteristics of the cinema display such that a satisfactory grading achieved on the grading reference display would 6 Which fundamentally they are; however, as we shall see in Section 32.8, a projector designed for the cinema is likely to include a transform to complement the DCDM perceptibly uniform encoding, and a flat-panel display may contain a gamma circuit to complement signals originating in a television or photographic environment. 7 The term ODT is retained in order to provide consistency with Figure 32.2; however, as can be seen from the diagram, ODTs are often located at the input of the output device rather than in the post-processing operation.

Colour in Cinematic Production – The Academy Color Encoding System 581 produce a matching rendition in the cinema. However, such an approach could result in a situation where the industry became permanently tied to an increasingly obsolescent standard as improved colour gamut projectors were introduced into the system but the requirement to continue to support legacy equipment remained. To avoid this situation, the ideal RDD is defined by the ACES to complement the RICD, that is, as indicated above, ‘the RDD is an ideal output device with an unlimited color gamut and a dynamic range exceeding any current or anticipated real output device’; however, whereas a capture device is theoretically capable of capturing all colours, a three-primary display is theoretically incapable of displaying all colours. Nevertheless, since both the cinema projector and the reference display are provided with signals encoded in a colour space, the primaries of which have chromaticity coordinates which are located external to the spectrum locus, then irrespective of the content of the graded material on the reference display, and subject to appropriate transforms and compromises of any gamut mapping, that content will be matched in chromaticity terms on the cinema display. Thus, there is not the requirement to match the chromaticity gamuts of the cinema and grading reference displays, and both may independently be replaced by displays of wider gamuts, with only the secondary compromise of any gamut mapping required to be considered. In evaluating the reference display image, the colourist adjusts the grading controls to achieve the desired results; once adjustments are complete, the file is saved as the DSM. The contents of the DSM file are also transformed into the DCDM colour space in readiness for distribution; as will be described in the next chapter, the DCDM colour space is, like the ACES colour space, also based upon a colour gamut with chromaticity coordinates external to the spectrum locus and therefore does not impart any colour restrictions during encoding. Furthermore, the perceptible uniform encoding introduced by the DCDM encoder is precisely complemented by the cinema display ODT, making the distribution path transparent to colour information. The cinema display ODT transforms the DCDM colour space signal to a linear signal with RGB coding and contrast range mapping appropriate to the display and environment characteristics, ensuring the displayed image will represent an ideal perceptible match to the grading reference display. As the white points of the two displays are marginally dissimilar, there will be very small differences in the colorimetric match of the displays, subject, that is, to the cinema display chromaticity gamut being identical to or larger than the reference display. Should the cinema display chromaticity coordinates fall inside of the chromaticity gamut of the grading reference display, then there will be a requirement to provide colour gamut mapping (Frohlich et al., 2014) to accommodate those colours which fall outside the cinema display gamut. In practical terms, perhaps the nearest set of primaries to that of the theoretical RDD are those defined as the ideal display primaries in Section 9.2, together with a very similar slightly compromised set which were subsequently specified in the ultrahigh definition television (UHDTV) system described in Section 20.4. As the latter set is likely to become the standard for extended gamut displays, it is also likely to be adopted at some point for the grading and cinema displays. Such displays would be capable of accurately rendering nearly all the Pointer surface colours, and in the intermediate period before such displays are widely adopted for review rooms and the cinema projector, their gradual and uneven introduction would be achieved without significant compromise to the rendered image. In order to ensure however that a minimum chromaticity gamut is achieved in the cinema, the SMPTE have defined a reference projector specification based upon the colour characteristics of the xenon-illuminated digital light processing (DLP) projector in ‘SMPTE RP 431.2.2007

582 Colour Reproduction in Electronic Imaging Systems D-Cinema Quality – Reference Projector and Environment’. This recommendation specifically indicates that the reference gamut is the minimum gamut required and that extended gamuts may be used. At the time of writing, laser-based projectors with a colour gamut approaching the UHDTV Rec 2020 gamut are starting to be introduced. 0.7 0.6 520 530 540 550 560 570 510 580 590 600 0.5 500 SMPTE 431-2 610 620 630 640 660 700 EE white 0.4 490 Pointer Rec 2020 surface v′ colours 0.3 480 0.2 470 Ideal 0.1 460 450 0.0 440 400 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 u′ Figure 32.7 Display chromaticity gamuts. The gamuts of the displays discussed are compared in Figure 32.7. As can be seen, the Rec 2020 gamut is very close to the ideal gamut, the latter being illustrated with a dotted line. The Rec 2020 gamut encompasses all but a few of the Pointer surface colours, and both gamuts significantly exceed the SMPTE 431-2 gamut based upon the xenon-illuminated DLP projector. 32.6.3 Which File for Archive? Whether the DSM or the file untransformed by the RRT is archived is a matter of some consideration, in as much as, at the time of writing, the characteristics of the RRT are not stable and it is not clear whether it will be defined in terms which enable it to be bidirectional. In philosophical terms, an archive file should be the closest representation possible to the original material, and as the ACES system has been defined to ensure that its colour space fully represents the original scene colours, then this is the version which should be archived.

Colour in Cinematic Production – The Academy Color Encoding System 583 However, if the RRT were defined in terms which enabled it to be used in either direction, then a case could be made for the convenience of storing the RRT version as the archive, on the basis that if the original coding was required, it could be obtained by applying the reverse RRT. A further point to consider is that as yet only candidate versions of the RRT have been published, so ‘baking in’ a transform for the archive, when in the future possibly improved versions of the RRT may become available, is a short sighted philosophy. 32.7 The Reference Rendering Transform The RRT is defined in the specification as: ‘the signal-processing transform that maps an ACES representation of an image to an Output Color Encoding Space (OCES) representation appropriate for viewing on the Reference Display Device’. However, although this defines its location in the workflow, it does not define either its purpose or its characteristics, and although a number of candidate versions of the RRT have been evolved by the IIF Subcommittee, none to date (2014) appear to have been judged to be entirely satisfactory by the post industry. To address the purpose and characteristics of the RRT in any detail is beyond the scope of this book; however, in fundamental terms, its purpose is to render the scene-referred captured image with its unlimited chromaticity gamut and contrast range in such a manner that, in conjunction with the appropriate ODT, it produces an accurately perceived and pleasing image on the defined display. Much of the difficulty in achieving an acceptable candidate RRT appears to be in establishing agreement on interpreting the subjective description of a ‘pleasing image’ and the manner in which such an image is achieved by the RRT. The premise (Fielding and Maier, 2009) appears to be that since the filmic appearance has been popularly accepted over many decades, then the rendition of the ‘pleasing’ picture should be based upon a generic version of the film ‘look’. In the cited article, the authors lay down a clear path of definitions associated with the RRT and provide a fundamental view of its required characteristics, based upon the film tone curves, to provide the pleasing rendered image. Work continues in this area and further papers on the topic are available (McElavain et al., 2012) and (Iwaki and Uchida, 2013) which describe RRTs with varying levels of complexity based upon a combination of 1D and 3D LUTS and chromaticity transform matrices. 32.8 The Reference Display and Review Room The ACES specification states: ‘The image interchange framework reference projector and associated viewing environment are equivalent to those defined in SMPTE RP 431-2-2007’. Since both projectors and other displays are used as the reference display in some colour grading suites, it will be assumed that the specifications in the SMPTE recommended practice apply to any display used for grading. The recommended practice also refers to SMPTE RP 431-1-2006, which defines the characteristics of the screen associated with the projector. RP 431 has an identical set of conditions to cover both the review room and the theatre or cinema; however, the tolerances associated with the review room are generally tighter than those specified for the cinema. In the following, emphasis is given to the review room, its display and the interfacing of the latter to the IIF.

584 Colour Reproduction in Electronic Imaging Systems 32.8.1 The Reference Display 32.8.1.1 Pixel Count Although digital cinema uses the same nomenclature as television to describe the sampling structure of the displayed picture, that is, in terms of multiples of ‘k’, in cinematic terms, the value of k in the horizontal direction is 1,024 compared with 960 for television. The minimum sampling structure for the reference display should be 2,048 horizontal and 1,080 vertical pixels, that is, 2k. 32.8.1.2 The Reference Display Colour Space RP 431 defines a minimum colour gamut for the reference display defined by the chromaticity coordinates of the display primaries shown in Table 32.4 but notes that, in practice, the reference display may have a larger gamut. The display white point represents the average values of the chromaticity coordinates of a number of xenon lamp sources whose characteristics change slightly with running time. Table 32.4 The chromaticity coordinates and white point of the reference display SMPTE RP 431-2 x y z u′ v′ Red 0.680 0.320 0.000 0.4964 0.5255 Green 0.265 0.690 0.045 0.0986 0.5777 Blue 0.150 0.060 0.790 0.1754 0.1579 White: average xenon 0.314 0.351 0.335 0.1908 0.4798 In early work associated with selecting the chromaticity coordinates of the cinema display, these primary chromaticities were referred to as the ‘P3’ set of primaries, and although that nomenclature was never, as far as the author is aware, used in any SMPTE or ACES specification, its use has become widespread within the media industry. The chromaticity gamut represented by these primaries is illustrated in Figure 32.7. All colours falling within the minimum colour gamut must be rendered within an accuracy of ΔE∗ab = 4. 32.8.1.3 Transfer Function RP 431 does not differentiate between the cinema projector and the review room display with regard to the transfer function parameters. As indicated in Section 32.5.2, the cinema projector is driven by a DCDM file which incorporates perceptibly uniform fixed-point encoding with an exponent of 1/2.6 and which RP 431 addresses by defining a projector complementary transfer characteristic based upon an exponent of 2.6. Since the projector itself is a linear device, this is achieved by including a suitable exponential transform in the projector. However, depending upon the actual IIF configuration adopted to serve the purpose of the simple configuration illustrated in Figure 32.6, it will often not be convenient to feed the review room display with a DCDM encoded signal. Furthermore, since other coding regimes have been proposed to service the IIF infrastructure (McElavain et al., 2012), it would complicate the required transforms unnecessarily to include

Colour in Cinematic Production – The Academy Color Encoding System 585 the 2.6 exponent transfer function component in the review room display device. Thus, although some post-operations may well include a reference display with a built-in transfer function to match the requirements of RP 431, together with the infrastructure to match, for the purposes of this description, it will be assumed that this is not the case. 32.8.1.4 Display Luminance The screen luminance is defined in RP 431-1 as 48 nits and a chromaticity of x = 0.314, y = 0.351. RP 431 also sets tolerances for these parameters and for areas away from the screen centre. 32.8.1.5 Sequential Contrast RP 431 defines sequential contrast to be measured to include the contribution of ambient light resulting from the review room environment defined in Section 32.8.2. The nominal value of the sequential contrast should be at least 2000:1. 32.8.1.6 Intra-frame Contrast Intra-frame contrast is measured using a 4 × 4 checkerboard pattern and should include the ambient light. It should have a minimum value of 100:1. 32.8.2 Review Room Environment 32.8.2.1 Ambient Light Level Stray light reflected from the screen should be minimized. The use of black, non-reflective surfaces on all surfaces other than the screen, along with recessed lighting, is recommended. With the display turned off, the ambient light level reflected by the screen should be less than 0.01 nits. 32.8.2.2 Reference Viewing Position for Colour Grading The reference viewing position for colour grading shall be at a distance of 1.5–3.5 screen heights. Lighting on work surfaces or consoles should be masked and filtered to eliminate any spill onto the screen. 32.9 The IIF Output Device Transforms (ODT) The ODTs serve the purpose of matching the generic OCES of the RRT to the colour spaces of the display devices or the distribution system. In general, this will require the ODT to provide two or more of the following transform elements: r Colour space transform matrix encoding r Perceptual uniform fixed-point r Gamma correction r Contrast law mapping

586 Colour Reproduction in Electronic Imaging Systems The OCES colour space is based upon the chromaticities of the ACES primaries, a linear encoding and a comparatively large contrast range. Thus, contrast law mapping will be required for all practical ODTs. At this time, there are no specific definitions for the characteristics of the ODTs, which are therefore likely to be proprietary in implementation; however, in general terms, they will follow the pattern described in the following examples. 32.9.1 The Reference Display ODT For the simplified IIF configuration illustrated in Figure 32.6, the reference display ODT would be a simple matrix to convert between the ACES (OCES) and the display colour space; in Worksheet 32(b), the required matrix is calculated for the chromaticities of both the RP 431 and the Rec 2020 sets of primaries and tabled in Tables 32.5 and 32.6, respectively. Table 32.5 Matrix for converting ACES to SMPTE 431-2 chromaticity space RACES GACES BACES R431 2.0877 −0.7413 −0.3464 G431 −0.1301 1.2304 −0.1003 B431 0.0067 −0.0638 1.0571 Table 32.6 Matrix for converting ACES to Rec 2020 chromaticity space R2020 RACES GACES BACES G2020 B2020 1.5091 −0.2590 −0.2501 −0.0776 1.1771 −0.0995 0.0021 −0.0311 1.0291 In reality, the current (2014) lack of a specified coding regime for transporting 16-bit digital signals around the IIF almost certainly means the OCES will require converting to a 12-bit digital encoding, which in turn will require some form of perceptible uniform fixed-point encoding in order to avoid contouring effects; see Section 13.8. In these circumstances, it will be necessary for the ODT to include LUTs to provide the complementary decoding characteristic in addition to the appropriate matrix. 32.9.2 The Digital Cinema Distribution Master ODT The DCDM colour space uses the CIE XYZ primaries and a perceptually uniform fixed point encoding which is based upon an exponential law with an exponent of 1/2.6. Thus, the DCDM ODT will comprise principally of a matrix with the characteristics shown in Table 32.7, which was calculated using Worksheet 32(b), followed by look-up tables to provide the exponential response.

Colour in Cinematic Production – The Academy Color Encoding System 587 Table 32.7 Matrix for converting from ACES to DCDM colour space RDCDM RACES GACES BACES GDCDM BDCDM 0.9999 0.0000 0.0001 0.3440 0.7282 −0.0721 0.0000 0.0000 1.0000 32.9.3 The Television ODT The current HD television system is defined by Rec 709, see Chapter 19, and to match this specification, the television ODT will require a conversion from ACES to the Rec 709 colour space which will include the appropriate gamma correction. The required matrix is calculated in Worksheet 32(b) and is detailed in Table 32.8. Table 32.8 Matrix for converting from ACES to Rec 709 colour space RACES GACES BACES R709 2.5513 −1.1195 −0.4318 G709 −0.2759 1.3660 −0.0902 B709 −0.0173 −0.1485 1.1658 Following matrixing and appropriate contrast law mapping, the gamma law to be applied to obtain the desired Rec 709 colour space is as defined in Section 19.3.2.1. 32.10 Colour Management in Production and Post 32.10.1 General Considerations The relative simplicity of the ACES system ensures that when fully implemented and properly managed, then when compared with those of photography, colour management problems are dramatically reduced. The definition of one colour space embracing all colours for capture and post, together with the detailed and mostly tight specifications for the grading/review environment, reduces the opportunity for mismanagement of colour to occur. 32.10.2 Potential Problem Area In specification terms, there appears to be only one area of comparative laxity and that is in specifying the field of view of the colourist. Since the only area of luminance in the field of view is the image, then the accommodation level of the eye will to a degree be dependent upon the area the displayed image subtends at the eye. Since this area is in turn dependent upon the viewing distance, which is specified as 1.5–3.5 screen heights, representing a change in field of view area of greater than 5:1, the level of adaptation between these two extremes could be significantly influenced, thus modifying the spatial dynamic contrast ratio of the eye (see Section 13.3.3). It is likely that this would result in a colourist located in turn at these

588 Colour Reproduction in Electronic Imaging Systems extreme positions applying different settings to the rendered image of those scenes containing significant areas of dark detail. 32.10.3 Preserving the Look One area of the production and post-operation which has traditionally been problematic is the preservation of the ‘look’ of the scene, set by the cinematographer during shooting or the early stages of post, when the material may be subject to subsequent grading on different equipment in different post houses. The problem in essence is one of communication; how can the settings arrived at during shooting be preserved and interpreted in the later stages of post? Grading equipment from different vendors often uses different nomenclature for the adjustments and often implements adjustments in a different mathematical manner. This problem was addressed by the expert cinematographers, colourists and engineers of the American Society of Cinematographers (ASC), who have evolved a simple but effective means of describing the adjustments made to the captured scene after notional corrections for colour balance (and the application of the camera IDT and the appropriate display ODT) and recording them in a colour decision list (CDL). Three mathematically defined adjustments are specified for each of the red, green and blue colour signals; these are defined as slope, offset and power. These unique names are given in order to differentiate them from the sometimes mathematical ambiguous names associated with the corresponding classic adjustments of gain, lift and gamma, respectively. Although these adjustments may be made available directly to the operator, the intention is that the usual proprietary adjustments are used to obtain the desired look and the vendor will interpret the adjusted values in terms of the ASC values to compile the CDL, where the ASC numbers are recorded for each scene. In addition to the original nine parameters defined above, a further parameter, saturation is provided as the tenth adjustment. The intent is that the CDL will be used by the colourist at the various grading stages to implement the previous adjustment decisions as the basis for the final adjustments which may then be ‘baked in’ to the signal. Clearly for this scheme to work successfully it is essential that all vendors implementing the scheme translate their adjustments to the ASC-defined parameters in the same manner. Fur- thermore, the set-up of the monitor and the viewing conditions associated with first adjustment must emulate as far as is practically possible the conditions of the grading suite.

33 Colour in the Cinema – The Digital Cinema System 33.1 Introduction This chapter describes the D-Cinema System, a system which is designed to ensure that the rendering of the image captured by the digital source master (DSM) in cinematic produc- tion is retained by the digital cinema distribution master (DCDM) during distribution and exhibition in possibly several thousand cinemas. This is the system instigated by the Digital Cinema Initiatives (DCI) organisation, an overview of which was given in Section 31.4.2 and which is specified in ‘SMPTE Standard 428-1 Digital Cinema Distribution Master – Image Characteristics’ (ST 428-1). The description of the system which follows is consistent with the approach adopted in describing the television and photographic systems in the preceding chapters; that is, the system is described from the bottom up. For a complementary description, where the system is described from the top down, which may lead to a more fully rounded understanding of the reasoning behind the adoption of some of the parameters and their values, the reader is recommended to the material by Maier cited in Chapter 31. These references also contain much fundamental material, which in this book is covered in earlier chapters. 33.2 System Requirements The DCDM encoding format was derived to be as far as practically possible device inde- pendent but also to acknowledge the requirements of the cinema projector and its viewing environment as first defined in Section 32.5.2 by the document ‘SMPTE RP 431-2 D-Cinema Quality – Reference Projector and Environment’ (RP 431-2). (It was noted in Section 15.2.1 that an output-referred system specification is dependent upon the parameters of the viewing environment.) Colour Reproduction in Electronic Imaging Systems: Photography, Television, Cinematography, First Edition. Michael S Tooms. © 2016 John Wiley & Sons, Ltd. Published 2016 by John Wiley & Sons, Ltd. Companion Website: www.wiley.com/go/toomscolour

590 Colour Reproduction in Electronic Imaging Systems The criteria for evolving the parameters and their values for the DCDM encoding format appear to have included the following: r A chromaticity gamut to encompass all colours of any foreseeable display technology r A colour space to encompass both the contrast range r and any foreseeable system white point chromaticity match within the tolerances r Constant luminance encoding r No perceived contouring of the image at any luminance level Capability of ensuring the rendered image in all cinemas shall r specified in RP 431-2 in absolute display terms, rather than in relative display terms, The encoding to be defined as is the case for television and photography The following description of the system, in describing how these criteria were met, leads to the definition of the system which is formally captured as a specification in SMPTE ST 428-1. 33.3 Image Structure The image structure is based upon the cinema ‘K’ definition of pixel numbers described in Section 32.6, where 1K is 1024 horizontal pixels by 540 vertical pixels; pixels are square; that is, they have an aspect ratio of 1:1. Three operational levels have been specified as shown in Table 33.1. Table 33.1 D-Cinema image structure and operational levels Operational level K level Horizontal pixels Vertical pixels Frames rate 1 4 4,096 2,160 24 2 2 2,048 1,080 48 3 2 2,048 1,080 24 The number of pixels at each operational level shall not exceed the numbers shown in Table 33.1. 33.4 The D-Cinema Encoding Colour Space 33.4.1 The Chromaticity Gamut In order that the chromaticity gamut encompasses all colours, the system primaries will be located externally to the spectrum locus. A number of proposals were considered and finally a decision was made to adopt the CIE XYZ primaries as the location for the DCDM RGB primaries. Such an approach has the advantage that constant luminance encoding follows automatically.

Colour in the Cinema – The Digital Cinema System 591 G 1.0 0.9 530 540 520 550 0.8 510 0.7 y 0.6 560 570 500 580 0.5 0.4 EE white 590 0.3 490 600 610 620 640 700 0.2 DCDM 480 0.1 470 0.0 460 R B 400 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 x Figure 33.1 The DCDM chromaticity gamut on the x,y chromaticity diagram. The DCDM chromaticity gamut is illustrated on the CIE x,y chromaticity diagram in Figure 33.1. The RGB primaries are co-located with the XYZ primaries. Although there were originally concerns that, as much of the space enclosed by the gamut represented non-real colours, there would be an apparent inefficient use of the code bits, it transpired this was not a relevant issue. The chromaticity coordinates of the primaries are shown in Table 33.2, and the chromaticity gamut is also shown on the u′,v′ chromaticity diagram in Figure 33.2. Table 33.2 Chromaticity coordinates of the DCDM primaries CIE XYZ x y z u′ v′ Red 1.0000 0.0000 0.0000 4.0000 0.0000 Green 0.0000 1.0000 0.0000 0.0000 0.6000 Blue 0.0000 0.0000 1.0000 0.0000 0.0000 EE white 0.3333 0.3333 0.3333 0.2105 0.4737

592 Colour Reproduction in Electronic Imaging Systems G0.7 0.6 0.5 0.4 DCDM v' 0.3 0.2 B0.1 R 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0 u' Figure 33.2 The DCDM chromaticity gamut on the u′,v′ chromaticity diagram. Since the system primaries are collocated with the CIE XYZ primaries, that is, R = X, G = Y and B = Z, and are referred to in ST 428-1 as the XYZ primaries, we will adopt that notation for the remainder of this chapter. 33.4.2 The Luminance Dimension It may be recalled from Chapter 13 that ideally the luminance dimension of the colour space to be defined is dependent upon the contrast range of both the eye and the display if the ‘crushing’ of dark tones in the scene is to be avoided. RP 431-1 specifies the screen highlight luminance to be 48 nits and the ambient light reflected from the screen with the house lights off to be a maximum of 0.03 nits. However, a more realistic definition of screen black is ‘cinema black’, which is defined as the luminance level of the screen with the house lights turned off, the projector operational and fed with a signal code level of 0, that is, the addition of the residual light from the projector and the ambient light reflected from the screen. RP 431-2 defines a minimum sequential contrast ratio for the projector of 1,200:1. It is well recognised that the higher the contrast ratio achieved, the more realistic and pleasing is the perceived image, and in terms of defining the luminance dimension of the colour space, the specification calls for a contrast range of 5,000:1 to be accommodated; that is, an implied assumption that a cinema black luminance level of less than 0.01 nits may eventually be achieved. 33.4.3 The Digital Coding Requirement The relationships between the contrast ranges of the eye and the viewing environment, and the perceptibility of digital contouring artefacts, was explored in some detail in Sections 13.3 and 13.6, respectively, where it was shown that by adopting perceptible uniform encoding with appropriate parameter values, digital contouring artefacts could be made to be imperceptible. Worksheet 13(c) plots the perceptibility curves of the eye and the luminance changes of fixed-point perceptible coding curves for any quantisation bit range between 8 and 16 and any exponent value between 1 and 3 using the highlighted green parameters in the area shaded pink. Noting that in addition to the requirement to produce a quantisation curve below the human visual modulation threshold (HVMT), the other criterion is to select the minimum quantisation bits to minimise storage and transport capacity requirements. Manipulating the number of encoding bits and the exponent or gamma value in the worksheet leads to a compromise of 12 bits and an exponent value of 2.6, as illustrated in Figure 33.3. Thus, 12 bits will provide for 4,095 quantisation levels.

Colour in the Cinema – The Digital Cinema System 593 1 Modulation 0.1 Weber–Barten x 10 0.01 Weber–Barten 1% 0.001 Weber–Barten HVT Display contrast range 12 bit encoding, γ = 2.6 0.0001 0.001 0.01 0.1 1 10 100 1000 Display luminance (nits or cd/m2) Figure 33.3 Luminance ΔL curve for 12-bit encoding with an exponent of 2.6. In Figure 33.3, the straight line curve for a 12-bit encoded signal using a perceptible uniform fixed-point encoding exponent of 2.6 is shown to be below the HVT curve throughout the dynamic range of the system. Thus, such a system will not introduce any perceptible quantisation contouring. Adopting the parameters derived above and using the same format as that used for television and photography, the relative relationship between the X,Y,Z tristimulus signals and those encoded by the system is therefore given by: X′ = INT[4, 095 × (K × X)1∕2.6] and similarly for Y′ and Z′ (33.1) where INT is the nearest integer value and the value of K, representing the coding white point level, has yet to be defined. 33.4.4 Accommodating the System White Points It was noted in the system criteria listed at the beginning of this chapter that the system should accommodate any foreseeable system white chromaticities. The notional system white, as indicated in Table 33.2, is equal-energy white (EEW), the white normally complementing the CIE XYZ primaries and, by definition, the white obtained when normalised X = Y = Z = 1, and therefore when x = 0.3333, y = 0.3333 and z = 0.3333. Other whites, particularly those further towards blue or red than EEW on the chromaticity diagram, will, for the same luminance, have a value of Z or X exceeding the value 1.0, because of the fall-off in the

594 Colour Reproduction in Electronic Imaging Systems 1.0 0.9 530 540 520 500 0.8 510 0.7 0.6 560 570 500 y 580 0.5 EE white 590 0.4 600 0.3 490 610 620 640 700 0.2 480 0.1 470 460 0.0 400 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 x Figure 33.4 Plane of possible system whites points. luminous efficiency function of the eye away from the peak of the Y response. Illuminants redder than EEW are unlikely to be considered as system white points, but those illuminants towards blue on the CIE daylight locus are likely to fall into this category and the encoding system must be scaled to accommodate them. By selecting two values of X and Z in turn below a maximum of 1.0, whilst keeping the Y value at a level of 1.0, the resulting x,y values may be calculated and plotted on the CIE x,y chromaticity chart (Worksheet 33(b) WS Table 1). The resulting two lines may then be extrapolated in both directions to meet and to intercept the spectrum locus respectively, producing a plane of possible system white chromaticity values, for XYZ normalised values not exceeding the value of 1.0, as illustrated in Figure 33.4. In Figure 33.4, the plane of possible system white chromaticities, representing X, Z values which do not exceed 1.0, is illustrated in white. As anticipated from the rationale outlined above, this is the Y-dominated sector of the spectrum; both the X and Z sectors contain no system whites. Figure 33.5 illustrates the chromaticity plots of the required system whites on an expanded chromaticity scale. In order to scale the system to accommodate the required system whites, it is necessary to determine the normalised XYZ values for the range of system whites under consideration: r The ACES system white D65 r The Cinema system white, defined by RP 432-1 r The CIE-defined system whites between D50 and

Colour in the Cinema – The Digital Cinema System 595 0.37 0.36 D50 0.35 SMPTE 431-1 0.34 D55 y D60 (ACES) EE white 0.33 D65 Xenon 0.32 0.31 0.30 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.30 x Figure 33.5 Plot of system white points. These whites, together with their values after perceptible uniform encoding, that is, raised to the power of 1/2.6, are calculated in Worksheet 33(b) and shown in Table 33.3. Table 33.3 System whites and their X′Y′Z′ values for a constant luminance value of 1.0 System white x y z X Y Z X′ Y′ Z′ EEW 0.3333 0.3333 0.3333 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 SMPTE 431-1 0.3140 0.3510 0.3350 0.8946 1.0000 0.9544 0.9581 1.0000 0.9822 ACES 0.3217 0.3377 0.3406 0.9526 1.0000 1.0086 0.9815 1.0000 1.0033 D50 0.3457 0.3586 0.2957 0.9640 1.0000 0.8246 0.9860 1.0000 0.9285 D55 0.3325 0.3475 0.3200 0.9568 1.0000 0.9209 0.9832 1.0000 0.9688 D60 (ACES) 0.3217 0.3377 0.3406 0.9526 1.0000 1.0086 0.9815 1.0000 1.0033 D65 0.3128 0.3291 0.3581 0.9505 1.0000 1.0881 0.9807 1.0000 1.0330 From Table 33.3, it can be seen that D65 is the system white with the highest Z′ value of 1.033, a value which must be accommodated in the coding range of 0 − 4,095. Let us assume initially that the code value of 4,095 takes on the Z′ signal value of 1.033; since the Y value remains at 1.0, it would have a code value of 4,095/1.033 or 3,964. However, in order to ensure that the D65 system white point is well within the plane of available white points, the Y maximum code value is set to 3,960. The ratio of the maximum code value to the maximum Y code value is thus 4,095/3,960, a factor of 1.034, which when transferred to the linear signals by raising it to the power of 2.6 takes on a value of 1.0911. This is the factor by which the linear signals must be reduced in order to accommodate a D65 system white point within the coding range.

596 Colour Reproduction in Electronic Imaging Systems 0.37 0.36 D50 0.35 SMPTE 431-1 0.34 D55 y D60 (ACES) EE white 0.33 D65 0.32 0.31 0.30 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.30 x Figure 33.6 Plot of enhanced system white points. In Worksheet 33(b) WS Table 3, the chromaticity coordinates are calculated for the revised code levels, resulting in a broader plane of chromaticities with a maximum relative lumi- nance of 1.0 and which now embraces the D65 white point, as illustrated in Figure 33.6. When this scaling factor is applied to equation (33.1), the following set of equations will result: X′ = INT[4, 095 × (X∕1.0911)1∕2.6] and similarly for Y′ and Z′ (33.2) 33.4.5 The Absolute Encoding Equations A further system requirement is that the encoding equations should express the display colour in absolute terms rather than the traditional approach of expressing them in relative or normalised terms. Thus, the projector screen maximum luminance defined in RP 431-1 should be applied to equation (33.2); this is specified for the reference projector as a luminance L = 48 nits. Thus, applying the factor 1.095 to this luminance level provides an absolute factor of 52.37, leading to the following equation as defined in ST 428-1: X′ = [ × ( L × X )1∕2.6] and similarly for Y′ and Z′ (33.3) INT 4, 095 52.37 where L = 48.

Colour in the Cinema – The Digital Cinema System 597 Transposing this equation enables code levels to be equated directly with the absolute levels of X, Y, Z at the screen. ( )( X′ )2.6 (33.4) X = 52.37 × and similarly for Y′ and Z′ L 4, 095 33.4.6 Signal Excursion Code Levels Signal excursion code levels will be discussed in terms of the luminance signal, Y, and the white and black levels. White Level White has a level of 1.0, so substituting values in the parentheses of equation (33.3) leads to: ( 48 × 1 ) 52.37 = 0.9166 Thus, the code value of white on Y is: CVY′ = INT[4, 095 × 0.91661∕2.6] = 3, 960. This is to be expected since the factor 52.37 was chosen to make this so. Although the factor was chosen to leave head room for a change to a system white of D65; nevertheless, it also provides a small margin of head room for signal excursions which may extend beyond white. Black Level As explained in Chapter 13, black level has many meanings depending upon the level to which the eye is accommodated, which is in itself dependent upon the average display luminance, the luminance of the relatively small area at the centre of attention and the luminance of the surrounding surfaces. Nevertheless, it was indicated earlier in this section that the system should accommodate a contrast range of 5,000:1, which if used would provide a subjective black level represented by a level of 0.0002 of white level. Using equation (33.3) to obtain the code value for this black level leads to a code value of 362; since this level is regarded as approaching the limit of perception, there is clearly significant foot room for excursions below this level. Absolute or Relative Colorimetric Encoding It was noted that the maximum level of cinema black is related to the specified minimum cinema contrast range of 1,200:1, that is, a level of 0.00083, well above the notional limit of 0.0002 noted above; therefore, it is apparent from the discussion in Section 13.9, that in this case, the contrast range is limited by the cinema black level rather than the encoding parameters. Figures 33.7 and 33.8 illustrate the manner in which dark tones in the scene are compressed for the range of contrast ratios currently specified and those possibly achieved in the future. Interestingly, assuming a spatial dynamic contrast ratio of the eye of 5,000:1, a theoretical improvement in the contrast range from 5,000:1 to 10,000:1 will extend the perceivable range of dark tones, since the level of cinema black will effectively raise the otherwise unseen changes in projector dark tones to the level of perceptibility.

598 Colour Reproduction in Electronic Imaging Systems 1 Display relative luminance 0.1 0.01 0.001 1200:1 5000:1 ∞ 0.001 0.01 0.1 1 0.0001 0.0001 Display relative input level Figure 33.7 Effect of screen contrast range on tone reproduction. 0.01 Display relative luminance 0.001 1200:1 2000:1 5000:1 10000:1 ∞ 0.001 0.01 0.0001 Display relative input level 0.0001 Figure 33.8 Expanded version of Figure 33.7 If all cinemas had an identical cinema black level, it would be possible by adopting absolute colorimetric encoding to compensate for its existence by adjusting the dark tones at source to provide compensation but eventually a code level would be reached where changes were imperceptible and clipping would occur abruptly. A better approach is to allow the addition of the cinema black to the black represented by the dark tone code levels, ensuring there will always be a perceived change in dark tone level from the screen, albeit the change will be less

Colour in the Cinema – The Digital Cinema System 599 than that of an absolute colorimetric encoding system. The latter would provide larger, more accurate changes down to the cinema black level but would clip all signals below this level. In practice, the situation is resolved subjectively and satisfactorily by the colourist, who, being located in a very similar lighting environment to the cinema, adjusts the black level for subjectively optimum pictures. 33.5 DCDM Interfaces 33.5.1 The Input to the DCDM Encoding System By the nature of the digital cinema configuration, it would be inconvenient to undertake the DCDM encoding other than at the post-house following completion of the DSM. Thus, the DCDM encoding is likely to be incorporated within the image interchange framework (IIF) of the post-house as one of the output device transforms (ODTs), as described in Section 32.7.2. 33.5.2 The Output of the DCDM Encoding to the Projector Current projectors are fundamentally linear devices and thus require to be fed by linear signals; in consequence, each projector will require transform circuitry to be installed which matches the perceptually encoded DCDM signal to the characteristics of the projector. The transform circuitry comprises two stages: linearization of the perceptually uniform encoded signal and a transform matrix to match the encoded primaries to the primaries of the projector. The linearization will be accomplished by three look-up tables embodying the relationships: ( )( ) X = 52.37 × X′ 2.6 and similarly for Y′ and Z′ L 4, 095 Once in the linear domain, the signal will require 16-bit encoding to avoid the perception of contours resulting from digital encoding, as explained in Section 32.2. The parameters of the reference projector colour gamut are specified by RP 431-2 and are incorporated in Table 33.4; it will be noted that this is a full-colour space specification. Table 33.4 The SMPTE reference projector primaries chromaticities and colour gamut SMPTE 431-2 x yYz u′ v′ Red 0.6800 0.3200 10.1 0.0000 0.4964 0.5255 Green 0.2650 0.6900 34.6 0.0450 0.0986 0.5777 Blue 0.1500 0.0600 3.3 0.7900 0.1754 0.1579 White Av xenon 0.3140 0.3510 48.0 0.3350 0.1908 0.4798 The chromaticity gamuts of the CIE XYZ, RP 431-2 and Rec 2020 primaries are illustrated in Figure 33.9.

600 Colour Reproduction in Electronic Imaging Systems Figure 33.9 Chromaticity gamuts of cinema primaries. In Worksheet 33(a), the parameters of the transform matrix for matching the DCDM pri- maries to the reference projector primaries are calculated and the coefficients are shown in Table 33.5. Table 33.5 DCDM to reference projector primaries matrix XYZ RRP431-2 2.4381 −1.0180 −0.4201 GRP431-2 −0.7113 1.6897 0.0216 BRP431-2 1.0507 0.0369 −0.0876 As laser projectors are introduced, it will be necessary to provide an alternative transform matrix to match the DCDM primaries to the primaries of the laser projector. In the event that the laser primaries match those specified by Rec 2020, as listed in Table 33.6 and illustrated in Figure 33.9, the matrix shown in Table 33.7 would be required.

Colour in the Cinema – The Digital Cinema System 601 Table 33.6 The chromaticity coordinates of the Rec 2020 primaries Rec 2020 x y z u′ v′ Red 0.7080 0.2920 0.0000 0.5566 0.5165 Green 0.1700 0.7970 0.0330 0.0556 0.5868 Blue 0.1310 0.0460 0.8230 0.1593 0.1258 White D65 0.3127 0.3290 0.3583 0.1978 0.4683 Table 33.7 DCDM to Rec 2020 primaries matrix XYZ RRec2020 1.6316 −0.3557 −0.2759 GRec2020 −0.6337 1.6165 0.0172 BRec2020 1.0260 0.0168 −0.0428 In terms of colour fidelity, the parameters of the digital production and cinema systems have been defined in such a manner that there is no theoretical limitation to perceiving a perfect rendition of the original scene. 33.6 Distribution Once the DCDM is complete, it is ready for packaging into the DCI digital cinema package (DCP), together with the other programme-related streams of data, such as the audio and subtitles, which are referred to generically as ‘essence’ data, where it is then in a format suitable for distribution to the cinemas. The description of the DCP is beyond the scope of this book.

34 Colour in Cinematography in the 2010s 34.1 Progress in Adopting the Digital Specifications The specifications for digital cinematography1 were published in the latter half of the 2000s, so mid-way through the 2010s is an appropriate time to review their status and the progress being made towards their adoption in the day-to-day working practices of the movie industry. The changes required in these working practices in moving from a film to a digital based operation were outlined in the Introduction to Part 5C, and are now virtually complete; the use of film for scene capture is now relatively rare. Furthermore, the demands for greater sophistication in high production value television programmes have led to these productions being shot on digital cine cameras and thus the requirement for many post-production or ‘post’ houses to accommodate both cine- and television-based productions in their workflows. The adoption of procedures encapsulating new specifications is dependent upon two princi- pal factors: the availability of the core equipment which incorporates them and the inclination of those employed in the craft and technology of the industry to adopt the procedures which embrace them. The first of these criteria, the availability of Academy Color Encoding System (ACES) compatible equipment in the workflow, has to a large extent been met. Over the period from the publishing of the ACES specifications in 2008, the vendors of equipment have responded by ensuring options are available to select ACES-based procedures. Digital cine cameras have been developed which provide the option of working to the specifications and camera vendors have provided the complementary input transforms to their camera spectral sensitivities to the vendors of grading equipment. Virtually all grading systems now incorporate a plethora of input transforms to cover the range of cameras employed in production and offer the option to the colourist of selecting from a number of colour spaces, which include the ACES colour space, the working colour space they prefer. The current candidate version of the reference 1 The Academy Color Encoding System (ACES) by the Science and Technology Council of the Academy of Motion Picture Arts and Sciences, and the Digital Cinema Initiatives (DCI) consortium system standardised by the SMPTE. Colour Reproduction in Electronic Imaging Systems: Photography, Television, Cinematography, First Edition. Michael S Tooms. © 2016 John Wiley & Sons, Ltd. Published 2016 by John Wiley & Sons, Ltd. Companion Website: www.wiley.com/go/toomscolour

604 Colour Reproduction in Electronic Imaging Systems rendering transform (RRT) and a number of output transforms are provided to match the working colour space to the colour spaces of the grading display and to the range of media formats. Grading displays may be either projector or LCD/LED based, and although the latter are very much improved in terms of overcoming the fundamental limitation of the technology to provide a contrast ratio matching the specification, nevertheless without some compromise in dynamic contrast response (see Section 8.3.1.3), there is still room for improvement. Laser projectors are being introduced with chromaticity gamuts approaching that of Rec 2020, in line with the option provided for in SMPTE ST 431-1-2006. The level of adoption of the ACES specifications into the working practices of production and post naturally depends upon the inclination of the individuals and companies involved to do so and it is apparent that large sections of the industry have not yet embraced the specifications in their day-to-day operations. Nevertheless, there is now a broad awareness of the trend to adopt the specifications and also of the working options available to adopt procedures based upon their use in the cameras and systems which support the production and post operations. The background to this situation will be explored further in Section 34.3. Since the publication of the ACES specifications in 2008, the Academy Subcommittee responsible has continued its work on evolving a generic RRT acceptable to broad sectors of the industry and a number of candidate versions have been released. The Subcommittee has also been receptive to feedback from industry regarding the adoption of its specifications. This work has led to a full review of the documentation and an augmentation of the specifications, which in December 2014 resulted in a re-issue of the supporting documentation. Where appropriate, this documentation has adopted a style designed to be less intimidating to those working in the field that do not have a specialisation in colour science. The augmented specifications and the elements of the new documents which pertain to colour reproduction are described in the next section and the ramifications of their adoption are further addressed in the section on systems and workflows (Section 34.3). Virtually all cinemas have now converted to digital projector–based screenings and are supported by the digital cinema package (DCP), which incorporates the JPEG 2000 compressed and encrypted version of the digital cinema distribution master (DCDM) standardised by the SMPTE and described in Chapter 33. 34.2 The ACES in the 2010s This section provides a description of the status of the ACES resulting from the publication2 in December 2014 of the package of documents which details the ACES specifications and describes their use. The Academy describes these documents as the first official release of version 1.0; the original documents covered in Chapter 32 are now described as ‘pre- release’ documents. Nevertheless, all the technical specifications appearing in the pre-release documents are incorporated in the current release. Excerpts from the ACES specifications are used with the permission of the Academy of Motion Picture Arts and Sciences. 2 https://github.com/ampas/aces-dev/tree/v1.0/documents/ZIP

Colour in Cinematography in the 2010s 605 34.2.1 A Brief Review of the December 2014 ACES Documentation 34.2.1.1 Overview The ACES Subcommittee of the Academy’s Science and Technology Council has undertaken a major review of the ACES and the level of its adoption in the cine industry, which has resulted in a clearer distinction between its work on the specifications and the work of the SMPTE on the standards which have evolved from these specifications. A consequence of which is that specifications which are ‘internal’ and relate to what might be seen as peripheral to the core ACES are retained as ACES specifications, whilst those which are of a more universal nature are adopted by the SMPTE and published as standards. In ACES terms, these latter specifications are described only in their technical bulletins. The full list of the current documents is shown in Table 34.1. Table 34.1 The ACES December 2014 documents Document Title Version/Date V2.0 Dec 2014 SPECIFICATION V1.0 Dec 2014 V1.0 Dec 2014 S-2013-001 ACESproxy, an Integer Log Encoding of ACES Image Data V1.0 Dec 2014 S-2014-002 Academy Color Encoding System – Versioning System V1.0 Dec 2014 V1.0 Dec 2014 S-2014-003 ACEScc, a Logarithmic Encoding of ACES Data for Use V1.0 Dec 2014 within Color Grading Systems V1.0 Dec 2014 V1.0 Dec 2014 TECHNICAL BULLETIN V1.0 Dec 2014 TB-2014-001 Academy Color Encoding System (ACES) Version 1.0 V1.0 Dec 2014 V1.0 Dec 2014 Documentation Guide TB-2014-002 Academy Color Encoding System Version 1.0 User Experience Guidelines TB-2014-004 Informative Notes on SMPTE ST 2065-1 – Academy Color Encoding Specification (ACES) TB-2014-005 Informative Notes on SMPTE ST 2065-2 – Academy Printing Density (APD) – Spectral Responsivities, Reference Measurement Device and Spectral Calculation and SMPTE ST 2065-3 Academy Density Exchange Encoding (ADX) – Encoding Academy Printing Density (APD) Values TB-2014-006 Informative Notes on SMPTE ST 2065-4 – ACES Image Container File Layout TB-3014-007 Informative Notes on SMPTE ST 268M:2003 Am1 – File Format for Digital Moving Picture Exchange (DPX) – Amendment 1 TB-2014-009 Academy Color Encoding System (ACES) Clip-level Metadata File Format Definition and Usage TB-2014-010 Design, Integration and Use of ACES Look Modification Transforms TB-2014-012 Academy Color Encoding System Version 1.0 Component Names

606 Colour Reproduction in Electronic Imaging Systems Where the documents in Table 34.1 relate to new or augmented ACES specifications, they are described in Section 34.2.2; the remainder of the documents fall broadly into three groups: r those that provide a more descriptive explanation of the specifications than was originally r provided; merely point to the corresponding SMPTE standard; r those that describe the augmented specifications. those that Those falling into these categories are briefly described in the following; the reader who requires a more extensive exposure of their contents is referred to the original documents. The contents and style of TB-2014-001 and TB-2014-002 particularly provide an indication of the importance the Subcommittee attaches to making the aims of the ACES known to a broader representative sector of the industry. It is difficult to improve upon the summary descriptions which appear in each of the technical bulletins; thus where appropriate, they have been extracted and placed in the descriptions which follow below, indicating as such by using italics for the extracted material. 34.2.1.2 TB-2014-001 In TB-2014-001, the application of the ACES to the production and post system is described as follows: ‘The key components of the ACES system are ACES encodings, ACES image files, ACES transforms and associated files, and an ACES clip-level metadata container that describes how the ACES image files were viewed when created or modified. ACES Version 1.0 is the first official release of these components. These components may be enhanced in subsequent releases based on industry requirements. Feedback from ACES Product Partners and end users made it clear that such a dynamic environment requires a clear system for version-control and naming of ACES components’. 34.2.1.3 TB-2014-002 In TB-2004-002 ‘User Experience Guidelines’, the policies of the committee are enunciated at some length, as the extract which follows indicates. The document provides very useful guidelines to those new to ACES and those who previously may have considered that because of the style of presentation, the effort required to understand the specification was beyond what they were prepared to commit: ‘A goal of ACES 1.0 is to enable widespread adoption by encouraging consistent imple- mentations in production and post-production tools throughout the complete film and television product ecosystem spanning capture to archiving. This is a very diverse set of tools, each used by professionals with different sets of skills. Furthermore, each man- ufacturer has established their own set of conventions for how to structure their user experience to best serve their market. Clearly, it is neither feasible nor appropriate for

Colour in Cinematography in the 2010s 607 these guidelines to specify in minute detail every aspect of a user interface (e.g. “all products must use a set of vertical drop-down menus labeled in 10-point Helvetica”). That said, the feedback from users on the first wave of products implementing the pre-release versions of ACES has been clear in the need for guidelines. One common comment is that the implementations are so different, figuring out how to configure ACES in one product is of little help when configuring the next. For example, naming conventions are different for no apparent reason. Another common concern is that the system is too reliant on acronyms and uses unfa- miliar concepts (e.g., what is a “reference rendering transform”?). Although some of these acronyms have become familiar within the inner circle of ACES product partners and early adopters, it must be acknowledged that the tolerance for these terms is much lower amongst the general population of industry professionals (e.g. how would one explain what an RRT is to an editor, CG animator, or anyone else without some color science background). As the ACES project transitions from technical development to wider industry deployment and the release of version 1, it is appropriate that we take a fresh look at how to portray the system to an audience that includes end-users in addition to engineers and color scientists. Although the technical terms and acronyms will continue to be used within the engineering community, these guidelines introduce a new set of terms intended to be simpler and more familiar to a wider set of users’. As an indication of the move away from acronyms where possible, the IDT is now referred to as the ‘input transform’ and the RRT and the ODT acting in series combination as the ‘output transform’, a transform which may be preceded by a user ‘look transform’. It is recommended that any reader intent upon a complete understanding of the means of implementing the ACES in his or her workflow should, following the completion of this chapter, access and read the Guidelines document particularly and review the advisability of accessing the other documents in the current release which are relevant to his or her needs. The Guidelines document also refers to the new version of the ACESproxy specification and the new ACEScc specification, which are described in Section 32.2.2 and referred to further in subsequent sections. 34.2.1.4 Technical Bulletins Which Relate Directly to SMPTE Specifications These technical bulletins primarily describe only the relationship between the original ACES specification and the almost identical SMPTE standard. TB-2014-004 includes the original (2008) ACES specification and additional informative material which explains how to calculate the values appearing in the table associated with the original Annex B, which lists the RGB values of the ColorChecker chart when taking into account the flare characteristic defined in the reference input capture device (RICD). TB-2014-005 describes the method for interfacing film captured images into and out of the ACES colour space.

608 Colour Reproduction in Electronic Imaging Systems TB-2014-006 basically refers only to the associated SMPTE ST 2065-4 standard for the ACES Image Container File Layout specification, which is intended to be compatible with software and hardware capable of reading and writing the OpenEXR format. TB-2014-007 basically refers only to SMPTE ST 268M:2003 Am1 – File Format for Digital Moving Picture Exchange (DPX) — Amendment 1, primarily for the purpose of specifying a container for images in the Academy Density Exchange Encoding (ADX). 34.2.1.5 TB-2014-009 TB-2014-009 specifies the ACES clip-level metadata file (‘ACESclip’), which is a ‘sidecar’ XML file intended to assist in configuring ACES viewing pipelines and to enable portability of ACES transforms in production. It is likely to become a useful tool for colour management and to minimise errors in the setting up of the workflow in post. 34.2.1.6 TB-2014-010 TB-2014-010 specifies the ‘Look Modification Transform (LMT) which imparts an image-wide creative “look” to the appearance of ACES images. It is a component of the ACES viewing pipeline that precedes the Reference Rendering Transform (RRT) and a selected Output Device Transform (ODT). LMTs exist because some color manipulations can be complex, and having a pre-set for a complex look makes a colorist’s work more efficient. In addition, emulation of traditional color reproduction methods such as the projection of film print requires complex interactions of colors that are better modeled in a systematic transform than by requiring a colorist to match “by eye”. The LMT is intended to supplement—not replace—a colorist’s traditional tools for grading and manipulating images’. The document ‘describes the use of ACES Look Modification Transforms (LMTs) for ACES- based color management. It provides several use cases for LMTs, defines how LMTs are expressed and are carried along with clips and projects, discusses LMT use in the context of a workflow employing ACES-based color management, and concludes with design guidelines for LMTs. This document also describes optimal use of LMTs and suggests several ways in which an LMT may be designed to support flexible mastering and archiving workflows’. 34.2.1.7 TB-2014-012 TB-2014-012 entitled ‘Component Names’ notes that ‘ACES component names have technical names that emerged from the engineering and development process. While the names make sense to the scientists, engineers and early adopters that “grew up” with the system, the larger adoption community targeted for adoption by ACES Version 1.0 does not have the historical knowledge and context of the ACES pioneers and a large majority of that community does not have the technical training needed to understand many of the existing names. This Technical Bulletin documents the ACES component naming conventions as agreed to by the ACES’. The new names listed in the document are shown in Table 34.2, together with, where appropriate, the original acronym as defined in Table 31.0, which appears in the introduction to Part 5C. Where appropriate, the recommended nomenclature will be used in the remainder of this chapter.

Colour in Cinematography in the 2010s 609 Table 34.2 Recommended nomenclature for ACES terms Original nomenclature Recommended nomenclature Colour primary sets ACES primaries 0 or AP0 SMPTE 2065-1:2012 primaries, a.k.a. ‘ACES primaries’ ACES primaries 1 or AP1 ACES ‘working space’ primaries, a.k.a. ‘Rec.2020+’ Encodings ACES2065-1 SMPTE 2065-1:2012, a.k.a. ‘ACES’ ACESproxy ‘ACES wire format’, a.k.a. ‘ACESproxy’,‘ACESproxy10’, ACEScc ‘ACESproxy12’ SMPTE 2065-1:2012 with Rec.2020+ primaries, log encoding, ACEScg floating point encoding, a.k.a. ‘ACES working space’ Input transform VFX-friendly encoding, i.e. integer version of ‘ACES working Look transform Output transform space’, with ACESproxy transfer function3 Transforms ACESclip file. Alternate: ACES xml Input device transform (IDT) Academy-ASC common LUT Look modification transform (LMT) ‘RRT plus ODT’ a.k.a. ‘ACES viewing transform’ format. Alternates: Common Containers LUT format, clf file Clip-level metadata file Academy-ASC common LUT format file, a.k.a. ‘CLF file’ 34.2.1.8 The Reference Rendering Transform Although the RRT is mentioned several times in the documentation, its only definition appears in Annex A to TB-2014-002 on ‘User Experience Guidelines’ as follows: ‘RRT (Reference Rendering Transform) — Converts the scene-referred ACES2065-1 colors into colorimetry for an idealized cinema projector with no dynamic range or gamut limitations’. Although a number of candidate versions of the RRT have been introduced since the publication of the original specification, 2014 marks the first official release of the RRT4 in versions for both the forward and the inverse directions. 34.2.2 The Augmented Specifications The December 2014 package of documentation includes a modified version of the ACESproxy specification, S-2013-001, and a new specification, S-2014-003, which describes a new working space for use in colour grading systems. In general terms both these colour space specifications share much in common, including a chromaticity gamut based upon a new set of primaries and a logarithmic encoding, albeit the characteristics of the logarithmic encodings are different. To differentiate the new primaries from the original ACES primaries, the original primaries are now designated as the AP0 primary set and the new primaries as the AP1 primary set. 3 It has been pointed out that this description is anomalous as ACEScg has a linear light transfer function and AP1 primaries, and is stored as a floating-point encoding. 4 https://github.com/ampas/aces-dev/tree/v1.0/transforms/ctl/rrt

610 Colour Reproduction in Electronic Imaging Systems 34.2.2.1 The AP1 Primaries and the ACES-AP1 Colour Space The AP1 primaries are based upon the Rec 2020 primaries, adjusted to slightly increase the saturation of each primary in order to minimise the risk of highly saturated Pointer surface colours falling outside of the resulting chromaticity gamut. The advantages of adopting the colour space based upon these primaries will be discussed in Section 34.3.3.3. Table 34.3 Chromaticity coordinates of the AP1 primaries xy z u′ v′ Red 0.7130 0.2930 −0.0060 0.5603 0.5181 Green 0.1650 0.8300 0.0050 0.0523 0.5914 Blue 0.1280 0.0440 0.8280 0.1565 0.1210 White D60 0.3217 0.3377 0.3407 0.2008 0.4742 The chromaticity coordinates of the AP1 primaries are shown in Table 34.3, together with the chromaticity of the system white point, which is D60, the same as that of the AP0 primaries set. 0.7 0.6 G 560 520 530 540 550 570 580 G590 510 600 610 620 63R0 640 660 700 0.5 500 R EE white 0.4 490 v′ 0.3 Pointer surface 480 colours 0.2 470 B 0.1 460 450 440 400 0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 u′ Figure 34.1 The AP1 primaries and the Rec 2020 primaries, the latter shown dotted.

Colour in Cinematography in the 2010s 611 The AP1 primaries gamut and the Rec 2020 primaries gamut for comparison are plotted in the chromaticity diagram derived in Worksheet 34(a) and illustrated in Figure 34.1. They are so similar that it is difficult to separate them on this full-scale chromaticity diagram. However, there are important differences: The red and green AP1 primaries are located just outside the spectrum locus, thus ensuring all maximum saturated red to yellowish-green hues are within the chromaticity gamut; similarly, the blue primary is also located just outside the spectrum locus, leading to slightly fewer saturated cyan hues being outside the chromaticity gamut than would otherwise be the case for the Rec 2020 primaries. References to the ACES colour space continue to be used without the use of an AP0 suffix but when it is transformed to the AP1 primaries it is identified here as the ACES-AP1 colour space. This colour space is the basis of both the ACESproxy and ACEScc colour spaces described below but confusingly, in the Academy specifications it is given a different name in each case, possibly because different groups defined these two colour spaces. The matrix required for the transform from ACES to ACES-AP1 is calculated in Worksheet 34(b) and shown in Table 34.4. Table 34.4 Matrix for transforming ACES signals to ACES-AP1 signals RAP1 RAP0 GAP0 BAP0 GAP1 BAP1 1.4514 −0.2365 −0.2149 −0.0766 1.1762 −0.0997 0.0083 −0.0060 0.9977 34.2.2.2 The S-2013-001 ACESproxy Colour Space As the ACES-AP1 colour encoding system uses a 16-bit half-precision floating-point colour encoding method, it is not directly compatible with current 10-bit and 12-bit transport systems; thus, the ACESproxy encoding system, which is required to transport the encoded image to a local monitor, specifies a 10-bit and a 12-bit encoding system, which as we have seen in Section 13.6.3 require a perceptible uniform encoding strategy if contouring artefacts are to be avoided. As logarithmic encoding is relatively common place and familiar, in digital cine cameras and grading systems, in order to emulate the characteristics of film, a logarithmic-type integer encoding is defined for the ACESproxy encoding system to transport a representation of the ACES-AP1 floating-point image. The contrast range of the ACES-AP1 signal is too large to be accommodated in the contrast range available to 10-bit and 12-bit fixed-point encoding systems; thus, it is limited by clipping the signal level below a point representing subjective black in the environment in which it is intended the proxy signal will be viewed. In consequence, the specification emphasises this encoding should not be used for any other purpose but on-set monitoring.

612 Colour Reproduction in Electronic Imaging Systems The transcoding formula in the specification is given in a slightly different and simplified form (ignoring the rounding rules) as: ACESproxy10 CV = 64 for ACES-AP1 CV ≤ 2−9.72 ACESproxy10 CV = (log2(ACES-AP1 CV) + 2.5) × 50 + 425 for ACES-AP1 CV > 2−9.72 (34.1) where CV = code value. (In S-2013-001 ACES-AP1 is referred to as ‘ACESproxyLin’) 800ACESproxy10 CV 700 600 500 400 300 200 100 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 ACES-AP1 Figure 34.2 The transcoding characteristic of the transform between ACES-AP1 and ACESproxy10. The characteristic for an exposure range one stop above a fully reflecting white surface used in the transcoding transform is plotted in Worksheet 34(c) and is illustrated in Figure 34.2. The code values on the y-axis result from a 10-bit coding. The constants in equation (34.1) are designed to place the (dark tones) clipped code values of ACES-AP1 in the range of code values defined in Rec 709, thus enabling established transport systems to be used for conveying the signal from, for example, a digital cine camera to an on-set monitor. If the monitor is set to white on a signal level of 0.9 (with the signal level of a perfect reflector being 1.0) and the clip level is set at 2−9.72, that is, a level of 0.001184, a contrast range of approximately 760:1 will result. Taking the log2 of 2−9.72 gives a figure of −9.72; thus, in equation (34.1), the addition of 2.5 provides a black level of −7.22; the multiplier of 50 provides the number of code values for each value produced by the log expression, and the addition of the 425 term places black level at the code value of 64, as is the case for Rec 709.

Colour in Cinematography in the 2010s 613 ACESproxy10 CV 1100 1000 940 0.001184 0.18 222.875 900 100.000 1,000.000 0.001 0.010 0.100 1.000 10.000 800 700 600 500 426 400 300 200 100 64 0 0.000 ACES-AP1 Figure 34.3 As Figure 34.2 but illustrated on a log/lin plot. Figure 34.2 is useful for illustrating the shape of the log characteristic, but to see the full contrast range of image values, it is necessary to use a logarithmic chart as illustrated in Figure 34.3. Some of the signal levels of interest and their code values are also shown. It will be noted that the characteristic requires an input of 222.875 to equate to the notional white code value of 940, ensuring the coding scheme will not clip the highlights of any realistic signal level. The specification also provides a formula for 12-bit coding which delivers the same shape of characteristic but the constants in the formula are changed to ensure that the black code level is at 256 and the white code value is at 3,760. It can be seen from an inspection of the graph in Figure 34.3 that the range of ACES-AP1 signal levels between 0 and 1.0 will be represented only by ACESproxy code values of between 64 and 550; even more crucially, the levels between 0.18 and white at 1.0 are constrained to a range of code values between 426 and 550, which means that when displayed on a standard Rec 709 monitor without an inverse transform in place, the images will be severely lacking in contrast but nevertheless useful for judging image composition. 34.2.2.3 The S-2014-003 ACEScc Colour Space Specification S-2014-003 is a new colour space introduced by ACES in recognition that in using its primary colour space (ACES-APO) for grading, it can in certain circumstances lead to problems and can be counter-intuitive in terms of the manner in which adjustment of the grading controls affects the appearance of the image.

614 Colour Reproduction in Electronic Imaging Systems In colorimetric terms, the ACEScc colour space is based upon the same ACES-AP1 colour space as that on which ACESproxy is based, as described in Section 34.2.2.2; however, the logarithmic encoding scheme is different in that it is defined to ensure there is generally no clipping of the ACES-AP1 contrast range, in order that following grading, the signal may be transformed back to the ACES colour space with no impairment. In certain circumstances, as we shall see in Section 34.3.3.3, there are occasionally exceptions to this general rule. Nevertheless, the ACEScc colour space is designed to be compatible with the ACESproxy colour space, particularly in terms of the use of on-set ‘look’ metadata using the American Soci- ety of Cinematographers (ASC) Colour Decision List (CDL), as described in Section 32.10.3. The use of the ACEScc is transient for grading purposes only, and in consequence, there is no file container specified, as it should not be used for interchange or archiving. The ACES signal is transformed to ACES-AP1 using the same matrix coefficients as those listed in Table 34.4 for ACESproxy. (In S-2014-003 ACES-AP1 is referred to as ‘ACESccLin’.) The ACES-AP1 signal is then encoded as 32-bit floating-point numbers as described in IEEE P754 using the formula in equation (34.2). ACEScc = (log2(2−15 × 0.5) + 9.72)∕17.52 for ACES-AP1 ≤ 0 (34.2) ACEScc = (log2(2−16 + ACES-AP1 × 0.5) + 9.72)∕17.52 for ACES-AP1 < 2−15 ACEScc = (log2(ACES-AP1) + 9.72)∕17.52 for ACES-AP1 ≥ 2−15 The transcoding characteristic resulting from equation (34.2) is plotted in Worksheet 34(c) and illustrated in Figure 34.4. 0.8 0.6 0.4 Scc 0.2 0.0 –0.2 –0.4 2.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 ACES-AP1 Figure 34.4 The transcoding characteristic ACES-AP1 to ACEScc. Although the log/lin graph in Figure 34.5 looks similar to the ACESproxy graph, it can be seen on inspection that the characteristic covers a much increased range of contrast.

Colour in Cinematography in the 2010s 615 1.2 1.0 0.8 Scc 0.6 0.4136 0.4 0.2 0.0 –0.2 0.001185 0.18 222.88 –0.4 0.00001 0.00010 0.00100 0.01000 0.10000 1.00000 10.00000 100.0000010,00.00000 0.00000 –0.6 ACES-AP1 Figure 34.5 The log/lin presentation of the same data portrayed in Figure 34.4. One drawback of the ACEScc scheme is that there will be circumstances when it cuts across the ACES philosophy of preserving all scene-referred colours; out-of-AP1-gamut colours will produce negative signals which cannot be encoded by a log characteristic as logarithms of negative numbers are not possible. In consequence, any negative values which would otherwise result from the ACES to the ACES-AP1 matrix will require mapping or clipping before the ACEScc logarithmic element of the transform. It is understood the Academy is addressing this issue. 34.3 Production and Post — System Configuration and Workflows 34.3.1 Introduction This final section differs from the remainder of the book in as much as it attempts to describe not so much the technology of colour reproduction but the manner in which that technology is being adopted within the cinematographic industries and how it may develop in the foreseeable future. In consequence, it is to a degree subjective; it is based to a large extent on contacts between the author and a small number of practitioners within the industry, both post house colourists and technologists concerned with the equipment that supports them. As indicated in the introduction to this part, the post industry in particular evolved a range of working practices during the change from film to digital operations, practices which were supported by manufacturers of grading systems who were prepared to provide different solutions for different post houses; this led to many different established patterns of work in the period before the Academy specifications were formulated. The post operation is complex and the successful practitioners within it are often reluctant to put aside their hard-learned individual procedures to adopt new ones based upon the

616 Colour Reproduction in Electronic Imaging Systems standards introduced in the late 2000s. Thus, although at the beginning of 2015, there is a well-understood basis of the advantages to be gained by using the ACES, and although it is conveniently available to use, its level of adoption is far from universal. Nevertheless, as a consequence of current post systems providing the capability of support- ing the ACES, there is a very comprehensive range of input and output transforms available to the colourist. So by default, the ACES underlying philosophical approach of a central work- ing space supported by appropriate input and output transforms has become the norm, albeit generally the ACES colour space is not yet used to provide the universal interface between elements of the system. However, this is a situation which is likely to change with the recent introduction of the ACEScc colour space, providing the colourist with a familiar logarithmic working space whilst generally retaining the advantages of scene-referred encoding. This section is split into two parts: the first part endeavours to describe in a very broad manner the current situation in terms of the practices of typical production and post operations, and the second part provides an example of how an ACES-based operation may look by the end of the current decade (2020). 34.3.2 Representative Current Post Operations The diverse requirements of post ensure there is not a ‘typical’ operation; the operation of each post house tends to reflect not only the market niche it was established to support but also the view of its principal colourists in specifying its equipment and its mode of operation. Thus, the descriptions which follow are merely representative; they do not purport to describe any particular operation. Figure 34.6 illustrates a post house grading room, configured in this instance for dealing with high end programmes and commercials, principally for television. When material for Figure 34.6 A post-production grading room.

Monitor Reference Rec 709/ Rec 709/ Rec 709 lin Rec 709 lin Rec 709 lin (Gamma 2.4) FPD Rec 709 lin FPD On-set Grading Colouris suite Grading controls Log/ Camera rec 709 imput Scene Camera transforms Grading capture & processo proprietary Working space Working log colour element transforms space (WCS) log or linear Digital cine cameras Post grading syste ∗The selected working colour space Figure 34.7 A representative 2015 prod

displays Cinema display DCDM/ DCDM/ RP 431-2 RP 431-2 RP 431-2 RP 431-2 projector projector st g s Cinema Movie output transform DCP HDTV RRT ACES/ Options 2 DCDM Options 2 g WCS*/ Output or DCDM transforms g WCS*/ em rec 709 Archive master file duction and post system configuration.

618 Colour Reproduction in Electronic Imaging Systems the cinema is being graded, the monitor is replaced by a projector, the surround lighting is extinguished and the level of the environmental lighting is brought down to very low levels. Typically, the monitor will be set to a highlight luminance of 100 nits and the viewing distance will be 2.5–3.0 times the picture height. 34.3.2.1 Representative System Configurations Figure 34.7 represents a range of representative configurations with options included to cover some of the more common approaches to current production and post workflows. Only those elements of the system directly associated with colour space choices are shown, and as can be seen, the colourist has a very wide range of transforms available, which can often lead to confusion in interpreting the correct colour transforms to be applied through the workflow; the input transforms are tinted pink and the output transforms are tinted green. Each vendor of digital cine cameras incorporates proprietary logarithmic transforms of a slightly different characteristic within the camera, partly to emulate film characteristics and partly to enable the coded signal to be ingested into the post grading system without contour artefacts. A log to Rec 709 output transform may be provided in the camera for on-set monitor display of the captured scene, though sometimes for convenience, the log encoded signal is used directly to provide only a low contrast image for image framing and set arrangement. The camera manufacturers supply the vendors of the post grading systems with appropriate camera input transforms for each of their cameras in order that the colourist has the correct transform to hand when ingesting new material. In addition, a number of other working colour spaces are made available to suit the requirements of the colourist. The grading system also offers a large number of output transforms to support the reference displays, the DCDM,5 the archive file if relevant and other various media markets. 34.3.2.2 A Representative Workflow The arrangement of displays and control surfaces in the grading position tends to be dependent upon the preferences of the colourists who make up the post grading team and whether the suite is being used for the grading of material for cinema or television. Two different approaches to the layout of the grading area are illustrated in Figure 34.8 and 34.9. The first job in post is to ingest the material from the camera and often, as the material for a project derives from different cameras with different log characteristics, some post houses or colourists will arrange for the input transforms to provide a common working colour space (WCS) for ingested material prior to the commencement of the grading operation. This maybe either a preferred camera colour space or sometimes a non-dedicated colour space such as, for example, the wide gamut Cineon-log working space. These selections are made from the very comprehensive range of features available on one or more of the grading system display panels, a typical example of which is illustrated in Figure 34.10. 5 The DCDM colour space in this context refers to the colour space defined in Section 33.4, that is, a colour space comprised of an XYZ chromaticity gamut, a tone response characterised by a power law with an exponent of 1/2.6 and a contrast range defined by 12-bit perceptible uniform encoding. Within the post operation, this colour space is sometimes referred to as the DCI XYZ colour space.

Colour in Cinematography in the 2010s 619 Figure 34.8 A monitor grading view. The configuration of the colour space transform elements following the working space varies considerably depending upon the practices of the post house and the individual colourist and whether grading is being undertaken for the cinema or DVD/television. These variations in configuration are accommodated by the ‘Option’ connections in the Figure 34.7. In the following, initially the ideal ‘default’ configuration is described, which uses colour management practices to match the working colour space to the various colour spaces of the Figure 34.9 A grading view in a high end post set-up position; the projection screen is not shown.

620 Colour Reproduction in Electronic Imaging Systems Figure 34.10 The detail in a typical grading set-up screen. environment. In the subsequent description, the alternative custom and practice approach still used by many in the industry is outlined. The Colour-Managed Workflow With reference to Figure 34.7, the output from the grading processor is in the form of the WCS, which in turn will be the favoured working space of the colourist, usually one of the logarithmic working spaces. This working space will be transformed to match the reference display working space, either the projector or the flat-panel display (FPD). Appropriate transforms are available to undertake this match, labelled as the WCS/DCDM or the WCS/Rec709, respectively. In this situation, the Options 1 and 2 selections will be connected to the output of the transforms. At the reference displays, the input transforms convert the characteristics of the signal to match that of the appropriate display. Thus, although there may be limited or no support of the ACES, this approach is based upon sound colour management principles. The path to the projector maybe configured to include the movie output transform, which includes the RRT and the ACES/DCDM transforms. The RRT has taken several years to stabilise and is not yet used universally within the post house environment. Despite the availability of the means to do so, indications are that few post houses have adopted a colour management approach to grading. The Custom and Practice Workflow It appears that the custom and practice evolved in the era of the digital intermediate process continues to be widespread in the post industry, that is, to use no matching transform between the logarithmic coding associated with the grading processor working colour space and the reference displays. In consequence, prior to the commencement of the grading operation, there will be a very significant mismatch of colour spaces in the workflow, which will be manifest by a very low contrast, low saturation rendered image on the reference display. The colourist compensates for this mismatch by using the grading controls to increase the contrast and saturation until a pleasing and acceptable image is rendered on the reference

Colour in Cinematography in the 2010s 621 0.8 0.6 Scc 0.4 Scc 0.2 Scc2.4 0.0 –0.2 –0.4 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 ACES-AP1 Figure 34.11 Rendering characteristic of a ‘custom and practice’ workflow on linear plot. display. However, it is unlikely that this compensatory adjustment is capable of achieving the same results as that achieved by a colour-managed workflow; the wide colour gamut of the working colour space will not match the display primaries, and it is unlikely that the tone adjustments will mimic the precise characteristics of the missing colour-managing transform. Effectively, the grading processor controls are compensating for the combined mismatched characteristics of the logarithmic working space and the reference display characteristics, in the case of the FPD, the emulated display gamma of 2.4. In Worksheet 34(c), a combined characteristic based upon a representative ACEScc logarithmic characteristic and the gamma of a Rec 709 display is calculated and plotted on to both linear and log/lin graphs, which are illustrated in Figures 34.11 and 34.12, respectively. Although the display gamma compensates to a degree for the high gain of the logarithmic characteristic at low signal levels, providing an improved point gamma, nevertheless, the shape of the resulting curves clearly shows why the rendered display is so lacking in contrast; for a signal input level from the camera equating to 1.0, the combined characteristic will produce a relative light output from the display of only 0.24. In addition, as was shown in Section 29.5.3.3, rendering a large chromaticity gamut signal onto a display with a smaller chromaticity gamut will lead inevitably to a desaturated display. It is to be hoped that the new initiatives from the Academy will precipitate a stronger movement towards embracing a colour management approach to the grading operation. Some of the major post houses configure display rendering transforms (DRT), which are effectively proprietary versions of the ACES RRT to convey both a particular ‘look’, which